Global-Scale Numerical Modeling of Seismic Wave Propagation Using PREM and Unstructured Triangular Meshing: Insights into the Layered Structure of Earth
نویسندگان
1 School of Mining Engineering, College of Engineering, University of Tehran, Tehran, Iran.
2 School of Mining Engineering, College of Engineering, University of Tehran, Tehran, Iran.
doi
10.22059/jesphys.2026.397888.1007701چکیده
The Preliminary Reference Earth Model (PREM) provides a robust framework for understanding seismic wave propagation through Earth’s layered interior. This study employs numerical forward modeling to simulate P-wave and S-wave ray paths, travel times, and apparent velocity curves within a 2D unstructured triangular mesh based on PREM. The mesh was optimized for numerical stability, achieving a mean element quality of 0.91 and facilitating accurate interpolation of PREM’s velocity profiles. Seismic responses are compared with those from simplified homogeneous and linear gradient velocity models to highlight the influence of Earth’s layered structure. Validation against an analytical homogeneous benchmark yielded a mean travel-time relative error of only %, while comparative analysis revealed that the linear gradient model deviates from PREM by as much as % at the core-mantle boundary. Results demonstrate that PREM’s velocity heterogeneities and discontinuities, such as the core-mantle boundary and liquid outer core, produce complex ray paths and variable apparent velocities, contrasting the straight paths and uniform velocities of the homogeneous model and the smoother trends of the gradient model. These findings underscore the necessity of detailed velocity models and advanced unstructured discretization for realistic seismic simulations. By providing a reproducible computational framework, this study affirms the effectiveness of advanced numerical techniques in capturing Earth’s internal dynamics. It emphasizes the critical role of layering in seismic wave behavior, offering insights into the development of next-generation Earth models.